Sunk Cost sunkcost.ai Data checked 2026-09-03

What hardware do you need to run Llama 3.1 8B Instruct?

Llama 3.1 8B Instruct at Q4_K_M is 4.9 GB of weights, with a context ceiling of 128k tokens. Fine for summaries and light chat. Don’t trust it with maths or multi-step tool use.

Cheapest machine that runs itMac mini M6, 16GB at $899
Shortest pay-backMac mini M6, 24GB — Pays back in 815 years
Fastest of the ones listedMac Studio M5 Ultra, 96GB — 98 tok/s at 32k context
Honest answer on costPays back in 842 years at 500k tokens a day.

How good is it, really?

On the Artificial Analysis Intelligence Index v4.3 it scores 7, which puts it in the Below every hosted tier band. Fine for simple, well-specified tasks. Noticeably less capable than anything the big labs sell today. Score source. See the whole table.

What it costs either way

Renting the same model costs $0.02 per million input tokens and $0.04 per million output (OpenRouter, cheapest active endpoint, checked 2026-09-03). Buying a machine only beats that if you use it hard enough, for long enough, that the hardware price divides down below the rental bill.

Machines that run it

MachinePriceSpeed at 32kPay-back
Mac mini M6, 16GB $899 12 tok/s estimated Pays back in 842 years Run the numbers
Strix Halo Framework Desktop, 32GB $1,269 21 tok/s estimated Pays back in 2,883 years Run the numbers
MacBook Air M5 (13-inch), 16GB $1,299 12 tok/s estimated Pays back in 1,217 years Run the numbers
MacBook Pro M5 (14-inch), 16GB $1,999 12 tok/s estimated Pays back in 1,873 years Run the numbers
Mac Studio M5 Max, 36GB $2,499 37 tok/s estimated Pays back in 1,394 years Run the numbers
DGX Spark GB10 Grace Blackwell, 128GB $4,699 22 tok/s estimated Pays back in 19,412 years Run the numbers

One machine per family, cheapest first. Speeds are measured where a public benchmark exists and estimated from memory bandwidth otherwise; the calculator says which for any configuration.

The specifics

Parameters
8B
Quantisation
Q4_K_M
Weights on disk
4.9 GB
KV cache
4.3 GB at 32k context
Maximum context
128k tokens (128k)
Licence
Llama 3.1 Community License
Sources
source 1, source 2